.

go/neo/neo_test.go

@@ -97,6 +97,73 @@ func (m *tMaster) Run(ctx context.Context) error {
 
 // ----------------------------------------
 
+/*
+func TestMasterStorage0(t0 *testing.T) {
+	t := NewTestCluster(t0, "abc1")
+	defer t.Stop()
+
+	M := t.NewMaster("m")
+	S := t.NewStorage("s")	//, "m:1")	// XXX do we need to provide Mlist here?
+	C := t.NewClient("c")
+
+
+	tM  := t.Checker("m.main")
+	tMS := t.Checker("m-s")
+	tSM := t.Checker("s-m")
+
+	// M starts listening
+	tM.Expect(netlisten("m:1"))
+	tM.Expect(δnode("m", "m:1", proto.MASTER, 1, proto.RUNNING, proto.IdTimeNone))
+	tM.Expect(clusterState("m", proto.ClusterRecovering))
+
+	// TODO create C; C tries connect to master - rejected ("not yet operational")
+
+	// S starts listening
+	tS.Expect(netlisten("s:1"))
+
+	// S connects M
+	tSM.Expect(netconnect("s:2", "m:2",  "m:1"))
+	tSM.Expect(conntx("s:2", "m:2", 1, &proto.RequestIdentification{
+		NodeType:	proto.STORAGE,
+		UUID:		0,
+		Address:	xnaddr("s:1"),
+		ClusterName:	"abc1",
+		IdTime:		proto.IdTimeNone,
+	}))
+
+	tM.Expect(δnode("m", "s:1", proto.STORAGE, 1, proto.PENDING, 0.01))
+
+	tSM.Expect(conntx("m:2", "s:2", 1, &proto.AcceptIdentification{
+		NodeType:	proto.MASTER,
+		MyUUID:		proto.UUID(proto.MASTER, 1),
+		NumPartitions:	1,
+		NumReplicas:	0,
+		YourUUID:	proto.UUID(proto.STORAGE, 1),
+	}))
+
+	// TODO test ID rejects (uuid already registered, ...)
+
+	// M starts recovery on S
+	tMS.Expect(conntx("m:2", "s:2", 0, &proto.Recovery{}))
+	tMS.Expect(conntx("s:2", "m:2", 0, &proto.AnswerRecovery{
+		// empty new node
+		PTid:		0,
+		BackupTid:	proto.INVALID_TID,
+		TruncateTid:	proto.INVALID_TID,
+	}))
+
+	tMS.Expect(conntx("m:2", "s:2", 2, &proto.AskPartitionTable{}))
+	tMS.Expect(conntx("s:2", "m:2", 2, &proto.AnswerPartitionTable{
+		PTid:		0,
+		RowList:	[]proto.RowInfo{},
+	}))
+
+	// M ready to start: new cluster, no in-progress S recovery
+	tM.Expect(masterStartReady("m", true))
+
+}
+*/
+
 // M drives cluster with 1 S & C through recovery -> verification -> service -> shutdown
 func TestMasterStorage(t *testing.T) {
 	rt	 := NewEventRouter()
@@ -117,16 +184,16 @@ func TestMasterStorage(t *testing.T) {
 
 	cM  := tracetest.NewSyncChan("m.main") // trace of events local to M
 	cS  := tracetest.NewSyncChan("s.main") // trace of events local to S XXX with cause root also on S
-//	cC  := tracetest.NewSyncChan("c.main")
+	cC  := tracetest.NewSyncChan("c.main")
 	cMS := tracetest.NewSyncChan("m-s")    // trace of events with cause root being m -> s send
 	cSM := tracetest.NewSyncChan("s-m")    // trace of events with cause root being s -> m send
-	cMC := tracetest.NewSyncChan("m-c")	   // ----//---- m -> c
+	cMC := tracetest.NewSyncChan("m-c")    // ----//---- m -> c
 	cCM := tracetest.NewSyncChan("c-m")    // ----//---- c -> m
 	cCS := tracetest.NewSyncChan("c-s")    // ----//---- c -> s
 
 	tM := tracetest.NewEventChecker(t, dispatch, cM)
 	tS := tracetest.NewEventChecker(t, dispatch, cS)
-//	tC := tracetest.NewEventChecker(t, dispatch, cC)	// XXX no need
+	tC := tracetest.NewEventChecker(t, dispatch, cC)
 	tMS := tracetest.NewEventChecker(t, dispatch, cMS)
 	tSM := tracetest.NewEventChecker(t, dispatch, cSM)
 	tMC := tracetest.NewEventChecker(t, dispatch, cMC)
@@ -139,8 +206,9 @@ func TestMasterStorage(t *testing.T) {
 	rt.BranchLink("s-m", cSM, cMS)
 	rt.BranchLink("c-m", cCM, cMC)
 	rt.BranchLink("c-s", cCS, rt.defaultq /* S never pushes to C */)
-	rt.BranchState("s", cMS) // state on S is controlled by M
-	rt.BranchState("c", cMC) // state on C is controlled by M
+//	rt.BranchState("s", cMS) // state on S is controlled by M
+//	rt.BranchState("c", cMC) // state on C is controlled by M
+	rt.BranchNode("c", cC)
 
 	// cluster nodes
 	M := tNewMaster("abc1", ":1", Mhost)
@@ -345,9 +413,9 @@ func TestMasterStorage(t *testing.T) {
 		},
 	}))
 
-	tMC.Expect(δnode("c", "m:1", proto.MASTER,  1, proto.RUNNING, proto.IdTimeNone))
-	tMC.Expect(δnode("c", "s:1", proto.STORAGE, 1, proto.RUNNING, 0.01))
-	tMC.Expect(δnode("c", "",    proto.CLIENT,  1, proto.RUNNING, 0.02))
+	tC.Expect(δnode("c", "m:1", proto.MASTER,  1, proto.RUNNING, proto.IdTimeNone))
+	tC.Expect(δnode("c", "s:1", proto.STORAGE, 1, proto.RUNNING, 0.01))
+	tC.Expect(δnode("c", "",    proto.CLIENT,  1, proto.RUNNING, 0.02))
 
 
 	// ----------------------------------------

go/neo/t_cluster_test.go

+// Copyright (C) 2017-2018  Nexedi SA and Contributors.
+//                          Kirill Smelkov <kirr@nexedi.com>
+//
+// This program is free software: you can Use, Study, Modify and Redistribute
+// it under the terms of the GNU General Public License version 3, or (at your
+// option) any later version, as published by the Free Software Foundation.
+//
+// You can also Link and Combine this program with other software covered by
+// the terms of any of the Free Software licenses or any of the Open Source
+// Initiative approved licenses and Convey the resulting work. Corresponding
+// source of such a combination shall include the source code for all other
+// software used.
+//
+// This program is distributed WITHOUT ANY WARRANTY; without even the implied
+// warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+//
+// See COPYING file for full licensing terms.
+// See https://www.nexedi.com/licensing for rationale and options.
+
+package neo
+// infrastructure for creating NEO test clusters.
+
+import (
+	"testing"
+
+	"lab.nexedi.com/kirr/go123/xnet/pipenet"
+)
+
+
+// TestCluster ... XXX
+type TestCluster struct {
+	name	string
+	net	*pipenet.Network	// XXX -> lo
+
+	gotracer	*TraceCollector		// XXX -> GoTracer
+	//tpy	*PyTracer
+
+	ttest	testing.TB	// original testing env this cluster was created at
+}
+
+// XXX stub
+type ITestMaster interface {}
+type ITestStorage interface {}
+type ITestClient interface {}
+
+// NewTestCluster creates new NEO test cluster.
+//
+// XXX ...
+//
+// XXX defer t.Stop()
+func NewTestCluster(ttest testing.TB, name string) *TestCluster {
+	return &TestCluster{
+		name:	name,
+		//...			XXX
+		ttest:	ttest,
+	}
+}
+
+// Stop stops the cluster.
+//
+// All processes of the cluster are stopped ...	XXX
+// XXX do we need error return?
+func (t *TestCluster) Stop() error {
+	//...	XXX
+	t.gotracer.Detach()
+	//XXX t.pytracer.Detach()
+
+
+	return nil
+}
+
+// NewMaster creates new master on node.
+//
+// The master will be accepting incoming connections at node:1.
+// The node must be not yet existing and will be dedicated to the created master fully.	XXX
+//
+// XXX error of creating py process?
+func (t *TestCluster) NewMaster(node string) ITestMaster {
+	//...	XXX
+
+	// XXX check name is unique host name - not already registered
+	// XXX set M clock to vclock.monotime
+
+	// tracetest.NewSyncChan("m.main")
+	// foreach node1,node2:
+	//	tracetest.NewChan("node1-node2")  // trace of events with cause root being n1 -> n2 send
+	//	tracetest.NewChan("node2-node1")  // trace of events with cause root being n2 -> n1 send
+
+	// for each created tracetest.Chan -> create tracetest.EventChecker
+
+	//rt.BranchNode("m", cM)
+	//rt.BranchState("m", 
+	//rt.BranchLink("n1-n2", ..., ...)
+
+	// XXX state on S,C is controlled by M:
+	// rt.BranchState("s", cMS)
+
+	return nil
+}
+
+func (t *TestCluster) NewStorage(node string) ITestStorage {
+	panic("TODO")
+}
+
+func (t *TestCluster) NewClient(node string) ITestClient {
+	panic("TODO")
+}

go/neo/t_events_test.go

@@ -154,6 +154,13 @@ type EventRouter struct {
 	byNode map[string /*host*/]*tracetest.SyncChan
 
 	// state on host changes. Takes precendece over byNode.
+	//
+	// XXX not needed? ( I was once considering state change events on C to
+	//	be routed to MC, because state change on C is due to M sends.
+	//	However everything is correct if we put those C state changes on to
+	//	byNode("C") and simply verify events on tMC and then tC in that order.
+	//	keeping events local to C on tC, not tMC helps TestCluster to
+	//	organize trace channels in uniform way )
 	byState map[string /*host*/]*tracetest.SyncChan
 
 	// event on a-b link
@@ -264,7 +271,7 @@ func (r *EventRouter) Route(event interface{}) (dst *tracetest.SyncChan) {
 			break // link not branched
 		}
 
-		// now as we ldst.a corresponds to who was dialer and ldst.b
+		// now as ldst.a corresponds to who was dialer and ldst.b
 		// corresponds to who was listener, we can route by ConnID.
 		// (see neo.newNodeLink for details)
 		if ev.ConnID % 2 == 1 {
@@ -315,6 +322,8 @@ func (r *EventRouter) BranchNode(host string, dst *tracetest.SyncChan) {
 }
 
 // BranchState branches events corresponding to state changes on host.
+//
+// XXX not needed?
 func (r *EventRouter) BranchState(host string, dst *tracetest.SyncChan) {
 	r.mu.Lock()
 	defer r.mu.Unlock()

go/neo/t_misc_test.go

@@ -18,6 +18,7 @@
 // See https://www.nexedi.com/licensing for rationale and options.
 
 package neo
+// misc testing utilities.
 
 import (
 	"context"

go/neo/t_tracepy_test.go

+// Copyright (C) 2018  Nexedi SA and Contributors.
+//                     Kirill Smelkov <kirr@nexedi.com>
+//
+// This program is free software: you can Use, Study, Modify and Redistribute
+// it under the terms of the GNU General Public License version 3, or (at your
+// option) any later version, as published by the Free Software Foundation.
+//
+// You can also Link and Combine this program with other software covered by
+// the terms of any of the Free Software licenses or any of the Open Source
+// Initiative approved licenses and Convey the resulting work. Corresponding
+// source of such a combination shall include the source code for all other
+// software used.
+//
+// This program is distributed WITHOUT ANY WARRANTY; without even the implied
+// warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+//
+// See COPYING file for full licensing terms.
+// See https://www.nexedi.com/licensing for rationale and options.
+
+package neo
+// NEO/py event tracer
+
+// TODO

go/xcommon/xtracing/tracetest/tracetest.go

@@ -349,6 +349,29 @@ func (d *EventDispatcher) Dispatch(event interface{}) {
 	outch := d.rt.Route(event)
 	// XXX if nil?
 
+	// TODO it is possible to emperically detect here if a test incorrectly
+	// decomposed its system into serial streams: consider unrelated to each
+	// other events A and B are incorrectly routed to the same channel. It
+	// could be so happenning that the order of checks on the test side is
+	// almost always correct and so the error is not visible. However
+	//
+	//	if we add delays to delivery of either A or B
+	//	and test both combinations
+	//
+	// we will for sure detect the error as, if A and B are indeed
+	// unrelated, one of the delay combination will result in events
+	// delivered to test in different to what it expects order.
+	//
+	// the time for delay could be taken as follows:
+	//
+	//	- run the test without delay; collect δt between events on particular stream
+	//	- take delay = max(δt)·10
+	//
+	// to make sure there is indeed no different orderings possible on the
+	// stream, rerun the test N(event-on-stream) times, and during i'th run
+	// delay i'th event.
+
+
 	// TODO timeout: deadlock? (print all-in-flight events on timout)
 	// XXX  or better ^^^ to do on receiver side?
 	//